AI & Claude Chat Vocabulary Guide

المخرجات التفاعلية (Artifacts)

Dedicated windows where Claude shows code, documents, SVGs or interactive web previews separately from the chat.

نموذج تصميم أولي

A visual draft or wireframe of a proposed interface or web page.

مساحة العمل المشتركة (Canvas)

A side-by-side workspace where user and AI edit a document or code together; ChatGPT and Gemini call it Canvas, in Claude the closest equivalent is Artifacts.

ماركداون (Markdown)

A lightweight plain-text syntax for headings, lists, tables, code blocks and bold/italic text.

كتلة الكود

A formatted, syntax-highlighted container that displays programming code clearly.

الموجِّه (الطلب)

The input — instructions, questions or context — that guides the AI's output.

موجِّه النظام

Foundational instructions set before the conversation that define the AI's persona, rules and boundaries.

الرمز (Token)

The basic unit of text a language model reads and writes — a word, part of a word or a character.

نافذة السياق

The maximum number of tokens a model can consider at once, including the conversation, files and its own reply.

سلسلة المحادثة

One continuous conversation, or a saved chat history.

إعادة التوليد

Asking the model for a new answer to the same prompt.

تعديل الرسالة وتفريع المحادثة

Editing an earlier message to start an alternative path of the conversation from that point.

رفع الملفات

Attaching documents, images or data files to a chat so the AI can read them.

التعمّق (الغوص العميق)

An extensive, detailed exploration of a concept or problem.

سلسلة التفكير

A technique that asks the model to reason step by step before giving the final answer.

التلقين بأمثلة قليلة

Including a few worked examples in the prompt to show the desired output.

التلقين دون أمثلة

Asking the AI to do a task without giving any examples.

تحديد الدور

Giving the AI a persona or expert role, e.g. "Act as a senior system architect".

الكود المُعاد هيكلته

Code restructured for clarity, efficiency and maintainability without changing what it does.

التحسين التكراري

Improving an output step by step through follow-up prompts.

المدخلات المحددة بفواصل

Using markers such as `"""`, `---` or XML tags like `<context>` to separate instructions from the text to process.

القيود

Explicit limits in a prompt, e.g. "under 100 words" or "valid JSON only".

الشخصية

The voice, tone and identity the model adopts in a conversation.

تسلسل الموجِّهات

Splitting a complex task into steps where each prompt uses the previous step's output.

وسوم XML

Named tags like `<instructions>` or `<example>` that structure a prompt; Claude follows them reliably.

التعبئة المسبقة للرد

Writing the first words of the assistant's reply through the API to steer its format, e.g. starting with `{` (support varies by model).

التلقين الفوقي

Asking the AI to write or improve a prompt for you.

الاتساق الذاتي

Generating several reasoning paths and keeping the most common answer.

هندسة السياق

Choosing which information, files, tools and history to give a model so it can do a task well.

المخرجات المنظمة

Requiring a fixed structure such as JSON, a table or a schema so software can reuse the result.

الأسئلة التوضيحية

Questions the AI asks before answering to fill in missing details.

الهلوسة

Confident but false, made-up or unsupported statements generated by a model.

الاستدلال (التوليد)

The live computation of producing output tokens from an input.

زمن الاستجابة

The delay between sending a prompt and receiving a response, often measured as time to first token (TTFT).

الإنتاجية

Generation speed, usually measured in tokens per second.

درجة الحرارة

A randomness setting: low gives focused, repeatable text; high gives more varied, creative text.

أخذ العينات النووي (Top-P)

Limits choices to the smallest set of likely tokens whose probabilities add up to P.

الضوابط الوقائية

Safety rules and filters that block harmful or unwanted outputs.

التعلم المعزز من ملاحظات البشر

Training that uses human ratings of answers to align model behaviour with human preferences.

الضبط الدقيق

Further training a pre-trained model on specialised data to improve it at a specific task.

التوليد المعزز بالاسترجاع

Connecting a model to external documents or a vector database so its answers are grounded in retrieved facts.

النموذج اللغوي الكبير

An AI model trained on very large amounts of text to understand and generate language.

التعلم العميق

Machine learning with multi-layer neural networks — the technology behind modern LLMs (not the same as a "deep dive").

الشبكة العصبية

Layers of connected mathematical units that learn patterns from data.

المحوِّل (Transformer)

The attention-based neural-network architecture behind most current LLMs.

آلية الانتباه

The mechanism that lets a model weigh which earlier tokens matter most when producing the next one.

المعاملات (الأوزان)

The learned numerical weights inside a model; model size is often stated in parameters.

بيانات التدريب

The text, code and other data a model learned from.

تاريخ انتهاء المعرفة

The date after which a model has no training data, so it may not know recent events without web search.

متعدد الوسائط

Able to work with more than text, such as images, PDFs or audio.

التضمينات المتجهية

Lists of numbers that represent the meaning of text, used for semantic search and similarity.

قاعدة البيانات المتجهية

A database that stores embeddings and quickly finds the most similar items.

Top-K

Limits each next-token choice to the K most likely options.

الحد الأقصى للرموز

A setting that caps how long the model's reply can be.

تسلسل الإيقاف

A string that makes the model stop generating when it appears.

البث التدريجي

Sending the reply token by token as it is generated instead of waiting for the full answer.

التفكير الممتد

A mode where the model spends extra tokens reasoning through a problem before answering.

نموذج مفتوح الأوزان

A model whose trained weights are published so anyone can run it on their own hardware.

المواءمة

Making AI systems behave helpfully, honestly and safely, in line with human intentions.

الذكاء الاصطناعي الدستوري

Anthropic's training approach in which a model learns to critique and revise its outputs against a written set of principles.

التحليل البنيوي

Reading input and breaking it into structured parts.

الاستخراج

Pulling specific fields — dates, amounts, names — out of unstructured text.

التلخيص

Condensing long documents, transcripts or code into the key points.

تلوين الشيفرة

Colour-coding keywords, strings and variables in code for readability.

القالب الجاهز

Standard, reusable code or text that needs little editing.

المخطط

A formal definition of data structure and field types, e.g. JSON Schema or SQL DDL.

الحالات الحدّية

Unusual or extreme inputs used to test that code still behaves correctly.

التعرّف الضوئي على الحروف

Extracting machine-readable text from images or scanned PDFs.

التنقية

Cleaning inputs or outputs by removing private details, credentials or malformed content.

محدِّد التنسيق

An instruction that fixes the exact output layout, e.g. "a Markdown table with 3 columns".

ملف CSV

A plain-text table format: each line is a row and commas separate the columns.

JSON

A lightweight text format for structured data using keys and values, common in API responses.

التجزئة

Splitting long documents into smaller pieces for processing or retrieval.

الاستشهادات

References pointing to the exact source passages an answer is based on.

البيانات الشخصية المعرِّفة

Information that identifies a person, such as name, phone, national ID or address.

تنظيف البيانات

Fixing duplicates, typos, wrong formats and missing values before analysis.

العصف الذهني

Generating many ideas, features or possible solutions.

تصحيح الأخطاء

Finding, isolating and fixing syntax or logic errors in code.

التدقيق اللغوي

Checking text for grammar, spelling, clarity and style.

الترجمة

Converting content between human languages, or code between programming languages.

التوليف

Combining information from several sources into one coherent explanation or document.

المخطط التفصيلي

A hierarchical structure of headings for a document, article or plan.

الشيفرة الزائفة

Plain-language steps that describe program logic before writing real code.

المعيار القياسي

A standard test used to compare quality, speed or efficiency.

التشغيل التجريبي الذهني

Walking through logic step by step without actually executing it.

التأسيس على المصادر

Restricting answers to verified, provided information so the AI does not assume beyond the facts.

حدّ المعدل

A cap on how many requests or tokens can be sent within a time window.

تتبّع الخطأ

The error message and stack trace shown when a program fails.

مراجعة الكود

Examining code for bugs, security issues and style before it is merged.

اختبار الوحدة

A small automated test that checks one function or component.

إعادة الصياغة

Rewriting text in different words while keeping its meaning.

ضبط النبرة

Rewriting text to sound more formal, friendly, concise or persuasive.

Claude

Anthropic's family of AI models and the assistant built on them, offered in tiers such as Opus, Sonnet and Haiku.

المشاريع

Workspaces in Claude that group chats with shared files and custom instructions.

المهارات (Skills)

Folders of instructions, scripts and resources that Claude loads when a task needs them.

بروتوكول سياق النموذج

An open standard introduced by Anthropic for connecting AI applications to tools and data sources.

الموصِّلات

Integrations that let Claude read from or act in apps such as Google Drive, Slack or Microsoft 365.

الذاكرة

A feature that lets the assistant remember relevant details across separate chats.

أنماط الرد

Preset or custom settings that change how Claude writes, e.g. concise, formal or explanatory.

وضع البحث

A mode where Claude runs multiple searches and reads sources before writing a cited answer.

Claude Code

Anthropic's agentic coding tool that works in your terminal and codebase.

ملف CLAUDE.md

A project file that Claude Code reads automatically for conventions, commands and context.

وضع التخطيط

A Claude Code mode where Claude researches and proposes a plan before changing any files.

Cowork

Claude's desktop mode for multi-step work on your files and tasks.

Claude في Chrome

A browser extension that lets Claude navigate and act on web pages.

التخزين المؤقت للموجِّهات

An API feature that reuses a repeated prompt prefix to reduce cost and latency.

واجهة الدفعات

Sending many requests to be processed asynchronously at a lower price.

حدود الاستخدام

Caps on how much a plan can be used within a time period.

الوكيل الذكي

An AI system that plans and takes multi-step actions with tools to reach a goal.

وكيلي (ذاتي التنفيذ)

Describes AI that acts on its own — searching, editing files, running code — rather than only replying.

استخدام الأدوات

The model calling external functions or APIs — search, calculator, database — during a task.

الوكيل الفرعي

A helper agent that a main agent launches to handle part of a task, often in parallel.

استخدام الحاسوب

An AI operating a computer screen through mouse and keyboard actions.

البحث في الويب

Letting the model search the internet for current information and cite it.

الواجهة البرمجية

A way for software to send requests to a service, e.g. sending prompts to a model from code.

حزمة تطوير البرمجيات

A library that makes calling an API easier in a given programming language.

الإنسان ضمن الحلقة

A workflow where a person reviews or approves AI actions before they take effect.

حقن الموجِّهات

Hidden instructions inside content — a web page, email or file — that try to hijack the AI.

كسر القيود

Attempts to trick a model into ignoring its safety rules.

التحيّز

Systematic unfairness in outputs inherited from data or design.

التقييمات

Test sets used to measure how well a model or prompt performs on a task.

التحقق من المعلومات

Verifying AI claims against reliable sources before relying on them.

خصوصية البيانات

Protecting personal or confidential information you share with AI tools.

المجاملة المفرطة

A model's tendency to agree with or flatter the user instead of being accurate.